Trademark Protection or Protectionism?∗
Eugenia Baroncelli †
Ekaterina Krivonos ‡
Marcelo Olarreaga §
Abstract
This paper explores the extent to which discrimination against foreign applicants inthe trademark registration process can be used as a “behind-the-border” barrier toimports. Prima-facie evidence shows that in some developing countries the ratio oftrademark registration to applications is much higher for national than for foreignapplicants, which is consistent with the notion of discrimination against foreign firms.A simple model is developed that suggests that incentives to discriminate are strongerwhen foreign firms produce products that are close in quality to the product producedby domestic firms. This hypothesis is then tested and empirically confirmed in threeof the four countries in our sample, suggesting that discretion and discrimination inthe trademark registration process can sometimes be used as a protectionist tool.
JEL classification numbers: O34, F12, F13, O57Keywords: Trademarks, Trade, Protectionism, Developing Countries.
∗This research was produced as part of a World Bank research program on Trade and Trademarks. Weare grateful to Maggie Chen, Carsten Fink, Beata Javorcik, Keith Maskus and Matthew Stern for helpfuldiscussions. We also wish to thank Judge Luo Dongchuan from the Supreme Court of the People Republicof China, and Catherine Chao and Zely Zhang from Duan & Duan Law firm for providing us with veryhelpful information on China’s trademark law. We are also grateful to Desmond Marumo, Registrar, andto Patrica Van Stavel, Executive Manager at CIPRO, South Africa, as well as to Andre Van Der MerweTrademark Director at DM Kisch and to Esme du Plessis, Patent Atorney at Adams and Adams. Ourthanks also go to Ethel Teljeur from TIPS in South Africa. The views expressed here are those of theauthors and should not be attributed to The World Bank or any of the institutions to which they areaffiliated.
†University of Bologna and The World Bank. e-mail: [email protected]‡University of Maryland and Inter-American Development Bank. e-mail: [email protected]§The World Bank, and CEPR. e-mail: [email protected]
Pub
lic D
iscl
osur
e A
utho
rized
Pub
lic D
iscl
osur
e A
utho
rized
Pub
lic D
iscl
osur
e A
utho
rized
Pub
lic D
iscl
osur
e A
utho
rized
Pub
lic D
iscl
osur
e A
utho
rized
Pub
lic D
iscl
osur
e A
utho
rized
Pub
lic D
iscl
osur
e A
utho
rized
Pub
lic D
iscl
osur
e A
utho
rized
1 Introduction
Dating from antiquity, craftsmen’s marks have been employed to identify the name of the
maker and prevent fraud. One of the many forms of intellectual property rights, trade-
marks are defined in the current economic and law literature as words, symbols or other
signifiers used to distinguish a good or service produced by one firm from the services or
goods produced by another firm (Landes and Posner 1987). Therefore, a “trademark” is
also “an element of a process of communication...which typically originated with the owner
or seller of a product and which is received by a prospective buyer of that product” (Papan-
dreou 2002). Tied to the dynamics of communication, a function of information is clearly
performed by a trademark, along with one of influencing, through the provision of such
information, the final choice of the prospective buyer toward the purchase of that specific
product.
Allowing economic agents to register trademarks, governments aim then at reducing
consumers’ search costs and, indirectly, at stimulating firms to increase or maintain the
quality and variety of standards of their trademarked products. Trademarks are powerful
instruments in reducing the informational asymmetries between producers and consumers.1
Trademarks help consumers to distinguish those quality features that are not observable at
the moment of the purchase of the product, such as the effectiveness of a shampoo or the
reliability of a hard disk. Faced with a choice between two apparently identical products, the
consumer would only have a 50 percent chance to pick the one that incorporates the desired
unobservable features. On the supply side, it would not be profitable for firms to incur
higher costs for (unobservable) quality improvements if these could not be signalled to the
prospective buyers to justify a higher sale price. In a market with information asymmetries,
without a collective enforcement of trademark rights, there would be no incentive for quality
improvements, the level of average quality would drop, and, at the extreme, the market for
high-quality products would disappear. If the buyer does not know the quality level of the
product she is about to purchase, but only the distribution of quality in the whole market,
she will only be willing to pay the price of the average-quality product. Expecting to be
offered only the price of an average-quality product, the sellers of above-average quality
products will soon drop off that market. If buyers are rational and anticipate this move
from the sellers of high-quality products, they may offer to pay an even lower price to
the remaining producers, which induces further exit from the group of producers of above-
average quality goods. This continues until high-quality goods are driven out of the market
and only the lowest quality good remains (Akerlof, 1970). It follows that, by protecting
trademark rights, public authorities secure the existence of markets for high-quality goods,
through the reduction of information asymmetries between sellers and buyers. Trademark
protection both helps reduce search costs for consumers and induces increase of quality
standards for firms. By limiting the use of a certain name to a particular firm, trademark
registration allows built-up of brand reputation. The brand with accumulated reputation
is then easily distinguished from other products. This distinction reduces the incentives
for companies with hit-and-run strategies to enter the market. Enforcement of trademarks
will reward firms that have a long-term business horizon, and care about establishing
reputation.2
A potential problem with trademark protection – as with any regulatory instrument– is
that it may be subject to political capture. By allowing certain firms to register their trade-
marks and not others, or by applying different standards to the enforcement of trademark
legislation, an important commercial advantage can be granted to some firms. This paper
explores the extent to which trademark registration discriminates against foreign firms – by
not granting (or delaying) their trademark registration – becoming an additional weapon in
the protectionist arsenal.3 By not granting (or delaying) trademark registration to foreign
producers, the trademark office can effectively shift profits from foreign to home producers.
The incentives to do so are explored in a situation with one domestic firm and a number
of foreign firms operating in the home market. The results suggest that the government
of the home country will have stronger incentives to discriminate against products similar
in quality to the ones produced by the domestic firm. The idea is that by not granting
registration to products of relatively similar quality the government is able to shift profits
to domestic firms without excessively hurting consumers.
The empirical part of this paper focuses on four developing countries, where a majority
2
of trademarks are held by non-resident firms (the reverse is true in high-income countries;
see Baroncelli, Fink and Javorcik, 2005). The four countries are China, Hong Kong, India
and South Africa.
2 Trademark Protectionism: Prima-facie Evidence
To assess the degree of discrimination against foreign applications in the area of trademark
registration, we constructed the following indicator of discrimination against foreign firms,
dc,i, located in country c and industry i trying to register a trademark in the domestic
market:
di,c =rHi /aH
i
rFi,c/a
Fi,c
(1)
where rHi is the number of trademarks registrations processed by the national trademarks
office in the name of residents (home producers) in sector i; aHi is the number of trademarks
applications filed directly with the national trademarks offices in the name of residents
(home producers) in sector i; rFi,c is the number of trademark registrations in sector i
received by national trademarks offices in the name of non-residents from country c plus
designations under either the Madrid Agreement or the Madrid Protocol (“which have not
been the subject of a refusal of protection or which are no longer open to such refusal”); and
aFi,c is the number of applications filed in sector i directly with the national trademarks office
in the name of non-residents from country c plus the number of trademarks designations
under the Madrid Agreement or Protocol.4
Our indicator is used as a proxy measure of the rate of transformation of domestic
applications into valid registrations, compared to the same ratio when the applicant is a
foreign individual or a foreign company (either a person that is not resident within the
territorial jurisdiction of the reporting country – hereafter referred to as “destination” –
or a company that has not been incorporated in the same territorial jurisdiction).5 Values
greater than one indicate that the rate of transformation is higher for domestic applicants
and that discrimination against foreign applicants may be present. A high variation in
this ratio across industries (even though values may be smaller than one) can also reflect
3
discrimination in some sectors against foreign producers.
In China and Hong Kong, for example, the manufacturing average ratio d is around
0.8 and 0.7 respectively6. All manufacturing sectors had an average d below 1, suggesting
that there is no evidence of discrimination against foreign firms trying to register their
trademarks. On the other hand in India and South Africa, the manufacturing average ratio
d is 1.3 and 1.5 respectively, suggesting that discrimination against foreign firms may be
present, as on average domestic applications are more likely to end up in registration.
There are 24 manufacturing sectors in South Africa that had a discrimination indicator
d above 1, and 20 sectors in India out of potentially 34 manufacturing sectors in WIPO’s
NICE classification.7 The discrimination indicator reached values above 2 for 4 manufac-
turing sectors in South Africa and 6 in India. The four sectors in which South Africa seems
a priori to discriminate the most against foreign firms in terms of trademark registration
are: ropes and strings (with d = 3.5), varnishes sector (d = 2.6), agriculture products n.e.c.
(including processed food, d = 2.2), and furniture and mirrors (d = 2.0). Other sectors
with a value of the discrimination index above 1 include musical instruments, common met-
als, hand tools and implements, vehicles, building material, textile, clothing and footwear,
meat fish and poultry, coffee, tea and cocoa, beers and soft drinks, alcoholic beverages, and
tobacco. The six sectors where the discrimination index takes values above 2 in India are:
firearms and ammunitions (d = 4.9), meat and fish products (d = 3.3), lace and embroi-
dery (d = 2.9), leather (d = 2.4), hand tools and implements (d = 2.3), carpets and mats
(d = 2.0). Other sectors with a value for the discrimination index above 1 in India include:
household and kitchen utensils, paints and varnishes, games and playthings, agricultural
and horticultural products n.e.c., coffee and tea, precious metals, common metals, textiles,
bleaching preparations, and apparatus for lighting.
There is also a significant variation in terms of discrimination across foreign source
countries applying for trademark registration in all four destination countries (see Figures
1 to 4). In China, Argentina, Greece and Israel face a discrimination index above 1.5,
indicating that the ratio of registration to applications is approximately 50 percent higher
for domestic applicants than for applicants from any of these countries. In Hong Kong,
4
applicants from Portugal and Russian Federation face an average discrimination ratio above
1.5. In India, Austria, Belgium, Canada, China, Finland and the Russian Federation are
the countries with discrimination ratio above 1.5. Finally, in South Africa applicants from
Argentina, China, Finland, Hungary, India, Korea, Luxembourg, Netherlands, Portugal
face a discrimination index above 1.5.
Thus, there seems to be prima-facie evidence that there is some degree of discrimination
in the registration process against foreign applicants in the four countries under examina-
tion. However, two points need to be raised. First, a value of the average discrimination
index below 1 does not mean there is no discrimination to be detected at all. For example,
if foreign firms were to have a better (less costly) application technology, discrimination
against foreign firms would be consistent with a value of d below 1. The cross-industry and
cross-country variation in the registration discrimination index, d, could provide important
information that could help us identify the presence or absence of discrimination against
foreign trademark applicants regardless of the average discrimination index. This is the
approach followed in the next section, where we provide an analytical framework which
allows us to identify incentives to discriminate against different country/sectors. Moreover,
the absence of discrimination in the registration process tell us very little about overall dis-
crimination in trademark regulation. The crucial source of discrimination could be present
at the level of enforcement of the trademark rights rather than at the stage of registration.
For instance, the fact that China seems to exhibit no discrimination in the registration pro-
cess is perfectly consistent with discrimination at the stage of enforcement of trademark
rights. Unfortunately, we have no data on enforcement and therefore a complete exercise
is not possible.
Second, one may wonder how countries can discriminate against foreign applicants if
different conventions and international agreements prevent them from doing so.8 A possible
answer is the excess discretion granted to trademark offices or the lack of clear rules for the
adjudication of trademarks. For example, the Chinese Trademark Law of 1983 (amended
in 1993) and the Implementing regulations set very few deadlines for either the Trademark
Office or the Trade Review and Adjudication Board to give feedback to private entities.
5
For example, after being notified of a refusal by the Trademark Office on grounds of non
conformity, such as identity or similarity with another (national) registered or preliminary
approved trademark, the (foreign) applicant has 15 days to apply to the Trademark Review
and Adjudication Board for a review. No deadlines are set either in the law (Art. 21) or in
the Implementation regulations (Rules 16 and 17) concerning when the Trademark Office
or the Board need to notify the applicant.
India had no provisions for well-known marks until the new Trademark Act was passed in
1999. This implies that owners of well-known foreign marks had no guarantee of having their
rights enforced under the Trade and Merchandise Act of 1958.9 Note that the data presented
above and used in the empirical section are for the period 1994-1998 (see the Appendix B)
and therefore correspond to regulations under the old trademark law. Another problem of
the old law that has now been remedied is the absence of an Appellate Board. However, the
extent of discretion granted to the Trademark Registrar continues to be noticeable in a few
areas. For example, regarding the examination of applications, Section 4 of Article 18 of the
Indian 1999 Trademark Act states that the Registrar is entitled to refuse an application or
to subject its validity to compliance of amendments, modifications, conditions or limitation
“if any, as he may think fit”. The new law also contains some regulations that discriminate
against foreigners when it comes to opposition to an advertised application. Article 21
grants discretion to the Registrar to ask for a security deposit to be provided in case the
opponent to the application is a foreign individual or a foreign firm that neither resides nor
carries its business within the territory of India. The law does not provide for any specific
sum or range of payment, so the discretion retained by the Registrar seems unlimited, as is
the scope of deterrence against a foreigner willing to secure her trademark rights in India.
3 Trademark protectionism: an analytical setup
This section develops an analytical model to explain discrimination in trademark registra-
tion against foreign firms. Some simplifying assumptions regarding firm behavior are made
in order to focus entirely on the government’s decision to use trademark protection as a
6
tool for shifting profits from foreign to domestic firms, without explicitly modeling strategic
interaction among firms. In Appendix A we relax some of these assumptions, by modeling
firms’ interaction explicitly, showing how this alters the results regarding discrimination in
the registration process.
We analyze a model of a vertically differentiated product with internationally segmented
markets. The good is produced by a domestic firm which competes with a number of foreign
producers in the domestic market (one from each country). There is a unique level of quality
q corresponding to each firm, such that qlow < q1... ≤ qd ≤ ... ≤ qhigh, where qlow is the
lowest available quality, qhigh is the highest quality and qd is the quality of the domestically
produced good.
Quality is not observable ex-ante by consumers which differ in their marginal valuation
of quality, denoted by a continuous and uniformly distributed variable θ ∈ [0, 1]. Each
consumer buys 1 unit of the good at the most, choosing the quality level that provides the
highest net utility which is defined as:
U(θ, q) = θq − p(q) (2)
where p(q) is the price of the good. Price is increasing and convex in q. Consumers
maximize their net utility given expected quality and their marginal valuation of quality.
Consumers with higher θ choose higher quality goods.10 Consumers with θ ∈ [0, p(qlow)qlow
]
do not purchase at all since it would give them negative net utility and the remaining
consumers chose the quality level of the good according to their θ.
In the absence of registered trademarks, consumers cannot observe the actual quality
level of the good. If the only registered trademark is the domestic one, all foreign brands
except the generic one with quality level qlow disappear from the market consistent with the
standard Lemons problem, as discussed in the introduction. The authorization to register
only the domestic good effectively cuts the market into three segments: consumers that
value high quality will buy the domestic brand and the remaining consumers will either
buy the generic (no trademark) foreign product or will not purchase the good at all. This
can be seen by focusing on a simplified consumer problem with one registered trademark
7
(domestic). Consumer’s problem becomes:
Max{θqd − p(qd), θqlow − p(qlow), 0} (3)
Consumers will prefer to buy the domestic brand if and only if:
θ ≥ p(qd) − p(qlow)
qd − qlow
(4)
Denote the critical level of willingness to pay for the domestic quality θ∗ = p(qd)−p(qlow)qd−qlow
. In
the absence of registered foreign trademarks the market share of the domestic firm is 1−
θ∗.
Allowing foreign firms to register trademarks reduces the informational asymmetry be-
tween producers and consumers, giving the latter more options to choose from. The market
becomes segmented corresponding to the number of brands registered on the domestic mar-
ket. Figure 5 illustrates the consumer’s optimal choice depending on her θ. Each curve
represents the net utility level associated with a certain quality level. If all foreign firms
are allowed to register their trademark products of quality levels q1 and qhigh, only con-
sumers with θ1 ≤ θ ≤ θ2 consume the domestic brand. If, on the other hand, goods of
quality q1 and qhigh are not registered trademarks, so that only the domestic brand and the
generic good are available, any consumer with θ ≥ θ∗ will purchase the domestic brand,
significantly increasing the market share of the domestic firm and its profits.
Discriminating against foreign firms can be optimal from a welfare perspective in the
same way as in a world with imperfect competition, positive tariffs may be optimal because
they shift profits away from foreign firms to domestic firms. Thus, the profit-shifting
argument is present in this setup and can justify the use of discrimination towards foreign
firms, which leads to an increase in domestic firms’ market share and profits. However, as
it can be seen in Figure 5, consumers lose from discrimination, since there are fewer choices
available. The loss of consumer surplus from discrimination against country 1 is illustrated
by the shaded area in Figure 5.
We assume that quality distribution is exogenous and that quality differentials are
8
constant throughout the entire quality range, such that q2−q1 = q3−q2 = ...qn−qn−1 = ∆ >
0, which significantly simplifies the derivation of welfare changes induced by trademarks
discrimination. Assuming endogenous quality makes the problem analytically intractable
once we allow for more than two quality levels (and we need at least four to illustrate
discrimination vis-a-vis countries with different quality levels).11
Producers face the same variable costs which are quadratic in quality, c(q) = q2. In order
to further abstract from strategic interactions among firms, we assume that each producer
earns the same per unit profit π.12 Thus, prices received by producers are predetermined
by the quality level and equal p(q) = c(q) + π. Total profits equal Πi = πDi, where Di is
the demand for the good produced by firm i.
The government agency (the Trademarks Office) has perfect information about the
quality levels of the goods originating in each country and controls trademark registration.
This is not a completely unreasonable assumption - trademarks authorities are more likely
to be better informed than consumers about quality levels of the goods, given that the
applications for trademark registration that they receive convey certain information about
the products and the firms. It is not unusual that the same officials who deal with intel-
lectual property rights (including trademarks) also handle matters of industrial property
protection (such as firm registration). Thus, the authorities’ insight about the products
of particular manufacturers would normally be superior to that of the consumers’. More-
over, if the country is a party to the Madrid System, trademarks officials have access to
additional information available through the WIPO Registry in Geneva.
The government maximizes social welfare, which is the sum of domestic profits and
consumer surplus:13
W = Π + CS (5)
The government maximizes W by choosing which firms are permitted to register trade-
marks. Note that by changing the variety of the registered foreign goods available to
the domestic consumers through trademark protection, the government affects the market
shares of all firms.
9
Consider a situation where the domestic good is of the lowest quality, as depicted in
Figure 6. First note that if all trademarks are registered initially, there is no point in
cancelling trademark registration of a foreign firm that produces a much higher quality
good than domestic firms. In terms of Figure 6, removing trademark registration for q3
or q2 when q1 is present has no impact on the market share of the domestic firm, which
continues to sell to consumers with pdqd
≤ θ ≤ θ1. At the same time, disappearance of
these brands unambiguously reduces consumer welfare, since those who previously bought
these goods are now left with fewer options to choose from. Total welfare is therefore
reduced as a result of the disappearance of these brands. Discriminating against country
1, on the other hand, increases the market share and the total profits of the domestic
firm. Although consumers lose in this case as well, total welfare may increase as a result
of discrimination. In the following we examine the conditions for positive welfare change
brought by discrimination against the closest competitor.
Starting from a situation when all countries are allowed to register their trademarks
on the domestic market, it is welfare-improving to discriminate against q1 if and only if
the extra profits created exceed the welfare losses to consumers. The change in consumer
surplus when q1 disappears from the market equals
δCS1 = −
θ3∫θ1
U(θ, q1)dθ −
θ2∫θ1
U(θ, qd)dθ +
θ3∫θ2
U(θ, q2)dθ
(6)
Since∫
U(θ, q)dθ = 12θ2q − θp(q) and the limits (as depicted in Figure 6) are θ1 =
p(q1)−p(qd)∆
, θ2 = p(q2)−p(qd)2∆
and θ3 = p(q2)−p(q1)∆
the expression simplifies to
δCS1 = −∆3 (7)
Thus, consumers unambiguously lose from discrimination against country 1. Note that
θ1 < θ2 < θ3 is insured by strict convexity of p(q), which implies 12p(qd) + 1
2p(q2) > p(q1).
The change in the domestic firm’s profits is
10
δΠ1 = π(θ2 − θ1) = π∆ (8)
Hence,
δW1 = ∆(π − ∆2)
The government decides to discriminate against country 1 only if ∆W1 > 0, which
requires π > ∆2. Assuming that π is high enough to induce discrimination, the government
does not allow q1 to be registered and the good disappears from the market. The next step
for the government is to decide whether discriminative action against country 2 should take
place. In this case the change in consumer surplus is
δCS2 = −
θ5∫θ2
U(θ, q2)dθ −
θ4∫θ2
U(θ, qd)dθ +
θ5∫θ4
U(θ, q3)dθ
(9)
where θ4 = p(q3)−p(qd)3∆
and θ5 = p(q3)−p(q2)∆
. Integrating and substituting θ2, θ4 and θ5 in
(9) simplifies the expression to
δCS2 = −3∆3 (10)
The domestic firm gains additional market share θ4 − θ2 and the additional profits are
the same as in the previous stage:
δΠ2 = π(θ4 − θ2) = π∆
The change in the total welfare from discrimination against q2 is then
δW2 = δΠ2 + δCS2 = π∆ − 3∆3 = ∆(π − 3∆2) (11)
The marginal increment in profits is constant as one discriminates against foreign firms
located further away in quality space, while the marginal loss in consumer surplus increases
with the quality distance between domestic and foreign firms. Thus, the change in social
11
welfare decreases as governments discriminate against foreign firms that manufacture goods
of very different quality than those domestically produced:
δW1 − δW2 = 2∆2 (12)
Since we assumed ∆ > 0, the expression is strictly positive, implying that the change in
welfare becomes smaller as the government moves from discriminating against country 1 to
discriminating against country 2 as well. Note that it now requires π > 3∆2 to discriminate.
While any π ∈ [∆2, 3∆2] would induce discrimination against country 1 (δW1 ≥ 0), further
discrimination would leave the country worse off (δW2 ≤ 0). With π > 3∆2 the government
will chose to discriminate in both cases, but the welfare increase is smaller in the second
stage. Note that welfare changes do not depend on the value of qd, but only on the quality
gap. The relative magnitudes of the welfare changes associated with discrimination against
countries 1 and 2 for different values of π and for quality gap ∆ = 0.1 are illustrated in
Figure 7.
Following the same approach, it can be shown that further discrimination (against a
hypothetical country 3) will produce a still smaller welfare change. It follows that, if initially
it is welfare-improving to discriminate, the incentives to do so diminish as we move further
away in quality space. The change in welfare continues to drop as more brands disappear
from the market, and it eventually becomes negative, when the loss in consumer welfare
outweighs the profits to domestic producers. At this point the trademarks office stops
discriminating and allows the higher quality goods to compete with the domestic brand.
It is straightforward to derive similar results for the case when the domestic good
is the highest in quality. In this situation the government also first chooses whether to
discriminate against the country which is closest in quality to the domestic brand and
then whether to discriminate against the next country. The conclusion is the same as
before: it may be welfare-improving to discriminate against the closest competitor, but
further discrimination would produce smaller welfare gains, which eventually would turn
into losses.
Thus, in this simple model a country discriminates against brands that are similar in
12
quality to the domestic one and allows trademark registration of those that are of a signifi-
cantly different quality. In other words, if trademark registration is used as a protectionist
tool, then discrimination is more likely in the case of foreign firms that produce goods
similar in quality to the domestic ones. This will be the basic hypothesis explored in the
empirical section of the paper.
4 Trademark Protectionism: Empirical Methodology
To test for the presence of protectionist rationale behind discrimination in trademark regis-
tration we explore the correlation between the registration discrimination index developed
in section 2 and a proxy for quality differences between domestic and foreign firms. More
specifically, for each of the four countries under examination (China, Hong Kong, India
and South Africa) we run the following regression:
di,c = β1∆qi,c + β2mc + αi + εi,c (13)
where di,c is the trademark registration discrimination index in industry i for products
originating in country c, ∆qi,c is the absolute value of the difference in quality in products
of industry i produced in country c versus products of the same industry produced in
the home country (China, Hong Kong, India or South Africa); αi is an industry dummy
included to capture any industry specific effect (e.g., higher trade protection in a particular
industry or better organized lobbies); mc is the share of imports from country c in total
imports of the destination country, and εi,c is an i.i.d error term. A negative β1 indicates
that as the difference in quality between domestic and foreign products increases, there
is less discrimination against foreign firms. This will be consistent with the notion of
trademark protectionism explored in the previous section. Import share mc is included to
test whether discrimination is more likely to occur in cases where the exporting country
already has a large share of the domestic consumption of foreign goods. This could happen,
if we assume that a large import share means that goods imported from that particular
country are similar in quality to the domestic goods (based on the Linder hypothesis). A
13
positive β2 would then strengthen the argument that discrimination is stronger in the cases
of close resemblance between foreign and domestic products.
The quality level of products in a particular industry i is captured by the share of sector
i’s exports to the QUAD (Canada, European Union, Japan and the United States) in total
industry exports. The difference in quality between products produced in the home country
and its trading partners is therefore calculated as ∆qi,c = |si,H − si,c|, where si,c is the share
of industry i’s exports to the QUAD in country c’s total exports of i. In the case of QUAD
members’ exporters, we also include their sales at home in the calculation of si,c. Subscript
H stands for the home country (i.e., destination country): China, Hong Kong, India and
South Africa. We propose taking the difference rather than the ratio when measuring ∆qi,c
in order to avoid losing observations when exports to the QUAD of country c are equal to
zero. Note however that estimates using the ratio are qualitatively identical to the ones
reported in Tables 1 and 2.
The basic assumption for using this ratio as an indicator of product similarity is that
products consumed in the QUAD are of relatively high quality, as QUAD consumers (i.e.,
consumers in rich countries) have higher θs than consumers in the rest of the world. This
hypothesis was first put forward by Linder (1961). Fink, Javorcik and Spatareanu (2003)
and Hallak (2003) have recently provided some empirical evidence in favor of this assump-
tion using very different approaches. Thus, if two countries have similar export shares to
rich countries, then it is likely that the products they manufacture are of similar quality.
Alternatively, the differences in quality levels could be captured by prices (or per unit
import values) of the domestic and the imported goods. An important caveat is that we
would need to compare very specific products. Our trademark protection data, however,
are aggregated at the industry levels – we can not calculate discrimination index for each
particular good. For example, for each exporting country we have discrimination index for
textiles and apparel sector, but comparing the quality of a ”representative” good in this
sector does not seem feasible, since we can only obtain per unit prices of specific goods,
such as men’s shirts or T-shirts.
14
5 Empirical Results
Table 1 shows the results of the estimation of equation (13) using a pool of the four
destination countries, with and without industry dummies (αi), with and without home
destination country dummies and with and without aggregate import shares from each
source country. All six regressions reported in Table 1 show a negative and significant rela-
tion between differences in product quality and discrimination against foreign firms in the
trademark registration process. The three last regressions also show that aggregate import
shares enter negatively, but insignificantly into the equation explaining discrimination in
the trademark registration process. Note that whenever we introduce import-shares we
use an instrumental variable estimator to control for the potential endogeneity of bilateral
imports to trademark discrimination. We instrument using geography variables (distance,
common border and common language) as in Frankel and Romer (1999). Results without
the use of instrumental variables are very similar, and qualitatively identical.
Table 2 provides estimates of equation (13) by country for each of our 4 destination
countries (China, Hong Kong, India and South Africa), but excluding the aggregate im-
port share variable mc. Three of the four countries in the sample show again a negative
and significant correlation between quality differences and discrimination against foreign
firms. The exception is China, where there was initially very little prima facie evidence
of discrimination against foreign firms. Note that in Hong Kong, whereas prima facie evi-
dence was also weak, the econometric evidence suggesting that the trademark registration
process can be used as a protectionist device is also very weak. It is negative and sta-
tistically significant before we introduce industry dummies, but it turns insignificant after
introducing industry dummies. Thus most of the discrimination could be explained for
example by industry lobbying (or other variables) rather than quality differences across
source countries in a particular industry.
Table 3 adds the (instrumented) import share variable, mc to the results provided in
Table 2.14 Again our indicator of quality difference, ∆qi,c is negative and statistically
significant in India and South Africa. In Hong Kong and China it is not statistically
significant. Note that this does not necessarily imply that there is no discrimination in Hong
15
Kong and China. As shown in the Appendix, if the weight given to domestic producers
in the government’s objective function is sufficiently large, then governments will want
to discriminate against all qualities, not only those that are similar to the ones that are
domestically produced. So one potential explanation of the results in Table 3 (and 2 for
that purpose) is that in China and Hong Kong the weight given to domestic producers is
sufficiently high, so that it makes sense to discriminate vis a vis all qualities and therefore
our variable ∆qi,c cannot capture this. An alternative explanation is that there is no
discrimination in the registration process.
However, this second explanation can be challenged by the coefficient of mc, which is
positive and significant in these two countries (China and Hong Kong). This implies that
the larger are imports from a particular country, the more likely is that country to be
discriminated against, which would also support our theoretical prediction if one assumes
that the Linder hypothesis holds (i.e., countries that produce and consume similar products
trade significantly with each other). It is insignificant in South Africa, indicating that most
of the discrimination is explained by quality differences at the industry/country level. In
India, there is evidence of discrimination both at the product level, as captured by ∆qi,c,
but also at the more aggregate level, as captured by mc.
We also performed two robustness checks. First we introduced tariffs as an additional
explanatory variable. The coefficient on tariffs is generally negative (except in the case of
China), but statistically insignificant. A negative coefficient would suggest substitutability
between tariffs and discrimination in the trademark registration process in terms of their
protectionist objectives. All other coefficients were qualitatively the same as in Table 3
(and Table 1 for the pooled regression). We also estimated the pooled and country specific
regressions using random effects and results were again very similar.15
6 Concluding Remarks
As traditional trade barriers, such as tariffs and quotas, have been eliminated in the de-
veloping world, much of the attention in the policy debate has shifted to the so-called
16
“behind-the-border” barriers to trade. Although it is difficult to give a comprehensive
definition of this concept, one could define it as a set of policies or an institutional setup
that explicitly or implicitly discriminates against foreign firms. The barrier explored in
this paper is the potential capacity of trademark offices to discriminate against foreign
firms in the registration of their trademarks. By not allowing foreign firms to register their
trademarks, these institutions can reduce the capacity of the foreign firms to penetrate the
home market.
Prima facie evidence for four developing countries suggests that there could be discrim-
ination in the registration process against foreign firms in some of these countries. A simple
model is then developed to show that discrimination is more likely to occur when products
offered by foreign firms are of similar quality to the ones produced by domestic firms. This
implication of the model is then tested for the four countries under examination (China,
Hong Kong, India and South Africa). Results are consistent with the alleged trademark
protectionism in at least two of the four countries: India and South Africa. It should be
kept in mind, however, that the absence of protectionism in the registration process is
consistent with the presence of protectionism at the level of enforcement. One potential
direction for future research is to explore the extent to which enforcement of trademark
legislation can also be used as a protectionist tool, in particular in countries where there is
little evidence of discrimination in the registration process.
17
References
[1] Akerlof, George A., ”The Market for ”Lemons”: Quality Uncertainty and the Market
Mechanism”, The Quarterly Journal of Economics 84 (1970):488-500.
[2] Baroncelli, Eugenia, Carsten Fink and Beata Smarzynska Javorcik, “The Global Dis-
tribution of Trademarks: some stylized facts”, World Economy 28 (2005):765-82.
[3] Fink, Carsten, Beata Smarzynska Javorcik and Mariana Spatareanu, “Income-related
biases in international trade: what do trademark registration data tell us?”, Policy
Research Working Paper #3150, The World Bank, 2003.
[4] Frankel, Jeffrey and David Romer, “Does trade cause growth?”, American Economic
Review 89 (1999):379-99.
[5] Hallak, Juan Carlos, “The effects of cross-country differences in product quality on the
direction of international trade”, RSIE discussion paper # 493, University of Michigan,
2003.
[6] Landes, William M. and Richard A. Posner, “Trademark Law: An Economic Perspec-
tive”, Journal of Law and Economics, 30 (1987):265-309
[7] Linder, Staffan, An essay on Trade and transformation, Uppsala: Almqvist & Wiksell,
1961.
[8] Nicita, Alessandro and Marcelo Olarreaga, “Trade and Production database, 1976-
1999”, available at www.worldbank.org/trade, 2001.
[9] Papandreou, A.G., “The Economic Effect of Trademarks”, California Law Review,
XLIV, 503-10, reprinted in: Ruth Towse and Rudi Holzhauer (eds.), The Economics
of Intellectual Property, Empirical Evidence, Trade Secrets and Trademarks, Vol III,
Northampton: Edward Elgar Publishing, Elgar Reference Collection, 2002.
18
Appendix A - The Theory
In this Appendix we investigate how the results presented in Section 3 change when prices
of the vertically differentiated goods are endogenously determined. In this setting firms’ per
unit profits vary with quality levels. The main result in section 3 holds (i.e., governments
have incentives to discriminate against foreign applicants with products of quality similar to
the domestic firm), but with some qualifications. First, note that once prices and markups
are endogenous, it only makes sense to discriminate if the government puts a sufficiently
high weight on producer surplus relative to consumer surplus in its objective function (for
political economy reasons, for example). Then governments will only discriminate against
similar quality products if the weight put on domestic producers is not too high. If the
weight is too high, then the government will have incentives to discriminate against all
products. Thus, our result in section 3 holds for a range of weights granted to profits in
the government’s objective function.
We analyze a model with four firms: one domestic and three foreign. The quality levels
are still exogenously determined, domestic quality being the lowest, and we maintain the
assumption of constant quality differentials, such that q1 = qd + ∆, q2 = qd + 2∆ and
q3 = qd + 3∆. Variable costs are still quadratic in quality, c(q) = q2. Profits of firm i are
then given by
Πi = (pi − qi)Di (14)
where Di is the demand for qi. Given the utility function (2), the range of marginal
19
valuation of quality θ ∈ [0, 1], and absent discrimination by the home country, the market
shares of the firms are:
Dd =p1qd − pd(qd + ∆)
∆qd
(15)
D1 =(p2 − 2p1 + pd)
∆(16)
D2 =(p3 − 2p2 + p1)
∆(17)
D3 =(∆ − p3 + p2)
∆(18)
Each firm maximizes its profits with respect to own price, taking the prices of the other
firms as given. The Nash equilibrium is then:
p∗d =qd(∆ + 39∆2 + 45q2
d + 76∆qd)
45qd + 52∆(19)
p∗1 =(qd + ∆)(2∆ + 78∆2 + 45q2
d + 100∆qd)
45qd + 52∆(20)
p∗2 =(8∆2 + 208∆3 + 45q3
d + 7∆qd + 375∆2qd + 220∆q2d)
45qd + 52∆(21)
p∗3 =(30∆2 + 338∆3 + 45q3
d + 26∆qd + 546∆2qd + 271∆q2d)
45qd + 52∆(22)
Social welfare is again defined as the sum of domestic profits and consumer surplus,
where profits are defined by (14) and consumer surplus is obtained by summing over all
consumers:
20
CS =∑
i
∫Di(θ)
U(θ, qi)dθ (23)
Discriminating against country 1 leaves three firms in the market instead of four. The
equilibrium prices change, leading to changes in market shares, total profits, consumer
surplus and social welfare. Profits of the domestic firm increase and consumer surplus
decreases as a result of discrimination.
Π1 − Π0 =1
2∆
A
(9qd + 20∆)2(45qd + 52∆)2(24)
CS1 − CS0 = −1
8∆
B
(9qd + 20∆)2(45qd + 52∆)2(25)
where both A and B are strictly positive.16 The change in total welfare is
W1 −W0 =1
8∆
4A −B
(9qd + 20∆)2(45qd + 52∆)2(26)
The expression is strictly negative in the range of parameters that ensures that each
firm sells in the no discrimination scenario.17 Therefore, it is never welfare improving to
discriminate even against the closest competitor. This conclusion is different from the
one obtained in Section 3 where all prices were predetermined. Note that the welfare
change is negative in this particular case where social welfare is a simple sum of producer
and consumer welfare. If, on the other hand, profits of the domestic firm weigh more in
the social welfare function than consumer surplus, for example due to industry lobbying,
21
discrimination can occur. In this case the social welfare is defined as:
W = kΠ + CS (27)
where k ≥ 1. This modification implies that the change in social welfare can be positive
(for sufficiently high values of k), which induces trademarks discrimination.
If the government discriminates against country 2 after the good produced by country
1 has disappeared from the market, the domestic firm only competes against country 3.
This move further increases the profits of the home firm, while reducing the welfare of the
consumers. At k = 1 the country on the whole is strictly worse off, but for higher values
of k discrimination against country 2 could lead to welfare improvements.
Since the final expressions for W1 − W0 and W2 − W1 are rather long and not easily
interpreted, the comparison of their magnitudes is made with the help of simulations. The
welfare changes associated with discrimination against country 1 (W1 − W0) and against
both countries 1 and 2 (W2 − W1) for a hypothetical case is depicted in Figure 8.
Whether or not the trademarks office decides to discriminate depends on the values of
k, qd and ∆. While at k = 1 discrimination always induces loss of social welfare, at higher
values of k discrimination may lead to welfare gains. Whether the change in welfare is
greater in the first stage (discrimination against country 1 only) or in the second stage
(discrimination against both countries 1 and 2) for different values of k depends on the
quality of the domestic good and the quality gap. Figure 8 illustrates the case where
qd = 0.1 and ∆ = 0.1. At k < 3.7 no discrimination would occur. If k ∈ [3.7, 4.9]
the welfare change is positive in the first stage, leading to discrimination against country
22
1, but not in the second stage, meaning that no discrimination action would be taken
against country 2. However, for values of k greater than 4.9 the government will choose to
discriminate against both countries.
Thus, relaxing the assumption of equal per unit profits for all firms produces a situation
where the decision to discriminate depends on the quality differential and on the weight
that the government attributes to the welfare of domestic producers.
Appendix B - The Data
The data set used for this study includes trademarks, trade, production and tariff data
for four countries (China, Hong Kong, India and South Africa), each of which is consid-
ered as the country of registration of brand names, and will hereafter be referred to as
“destination”). The data are at the industry level, covering the period from 1994 to 1998.
An average for this period is taken for every observation. The reason for this is that the
trademark registration process can often take more than one year and we therefore wanted
to avoid any biases due to the long delays that registration may often entail. In some of
the countries in our sample, the registration process can take easily two to three years.
Data on trademarks registrations and applications are disaggregated by country requesting
a registration (hereafter referred to as “source”).
The sources of the trademarks data are the CD-ROM version of the 1998 WIPO
database on trademarks, and the 2002 World Bank Trademarks Database (Baroncelli, Fink
and Smarzynska 2005), also based on WIPO data. The sector disaggregation used here is a
combination of the Nice Classification, the system used in both the WIPO and World Bank
23
sources, and the International Standard Industrial Classification (ISIC) at the three-digit
level, in which most of the output and trade data are reported. The final industry classi-
fication has 21 sectors. The country source disaggregation is the one provided by WIPO
and discussed in Baroncelli, Fink and Smarzynska (2005). There are potentially 40 source
countries in WIPO’s database; not all 40 countries have positive registration in our 4 desti-
nation countries (see Figures 1 to 4 for coverage in terms of source countries in each of our
destination countries). The trade, tariff and production data necessary to construct the
export shares in the calculation of ∆qi,c come from the World Bank Trade and Production
Database (Nicita and Olarreaga 2001). The data have been integrated with updated data
from the United Nations Statistics Comtrade database as well as with the United Nations
Industrial Development Organization (UNIDO).
24
Notes
1Trademarks are only one among the many tools through which firms can signal their products’ quality;
others being price and warranties.
2While a thorough analysis of the reasons for and against trademark protection is clearly beyond the
scope of this paper, it is worth indicating that the costs of drafting laws, maintaining a trademark register
and a registry, along with the administrative and judicial apparatuses necessary to deal with securing and
sanctioning trademark rights may not always be offset by the increase in both consumers’ and producers’
surpluses associated with trademark enforcement.
3Discrimination in the enforcement of trademarks can also potentially be used as a discriminating tool.
Note that, by explicitly doing so, the discriminating government would violate its national treatment
obligations in Article 2 of the Paris Convention (administered by WIPO) and therefore the WTO’s TRIPS
agreement. But agreements being incomplete contracts, there may be room for circumventing obligations.
4The data on registrations are from Baroncelli, Fink and Javorcik (2005). The data on applications are
from WIPO’s CD-ROM on trademarks. The Madrid Agreement of 1891 and the Madrid Protocol signed
in 1989 and entered into force in 1995 substantially reduce the transaction costs involved in registering
trademarks by allowing firms that reside in member states to file a single international application for
registration in multiple countries. For more on data sources and construction of variables, see the Appendix
B.
5One can imagine that foreign firms that are domestically established may also be subject to discrimi-
nation, or to less discrimination than an exporter without commercial presence in the host country. This
unfortunately cannot be analyzed with the data that is available to us. Applications by domestic residents
are likely to include applications of both national and foreign-owned firms based in the host country.
6Note that in Hong Kong all services industries have a ratio d above 1, probably indicating discrimination
25
in sectors where the Hong Kong economy is specialized.
7Although we do not have output data available for the NICE classification which includes 34 manufac-
turing sectors, we do have output data available at the ISIC 3 digit level of the UNIDO classification that
includes 29 manufacturing sectors. In all ISIC 3 digit industries, all four countries have positive output.
8Note that there is currently a trade dispute in the WTO regarding (potential) discrimination against
foreign applicants of trademarks (and geographical indications) regulations for agricultural products and
foodstuff in the European Union. The case was brought up by the United States and Australia and other
countries have requested to be third parties.
9Discretion in the interepretation of the law also explains why two restaurants in South Africa were
allowed to use the name “McDonald’s” after McDonald’s corporation missed a deadline to renew its regis-
tration in the early 1990s. It took multiple lawsuits and a reversal by the Supreme Court of South Africa
of an earlier decision by a lower court to McDonald’s to get the rights to its world famous name (case no.
547/95).
10Convexity of p(q) insures that the quality chosen is increasing in θ, since the solution to the consumer’s
optimization problem is θ = p′(q∗). It follows that ∂q∗
∂θ = 1∂θ
∂q∗= 1
p′′(q∗) ≥ 0
11However, it seems that the main analytical result of this paper (“governments will have incentives to
discriminate vis a vis foreign firms that produce similar products to the ones produced by domestic firms”)
will not be affected in a systematic way if we were to endogenize firm quality.
12This assumption is relaxed in Appendix A, and the basic result holds with some qualifications.
13Note that in this setup with constant marginal costs and no fixed costs, producer surplus and firm’s
profits are equal.
14These estimates may be inefficient if there is collinearity between mc and ∆qi,c. However the cor-
relation coefficient between these two variables is very low (it varies between 0.01 and 0.1 depending on
26
the destination country) and the variance inflation factor for these regressions oscillates between 1 and
4 suggesting that multicollinearity is not a problem. This is not necessarily surprising as ∆qi,c varies by
sector and country, whereas mc varies only by country.
15Results of these two estimations are available upon request.
16The final expressions for A and B in terms of qd and ∆ are too long to be reported here, but can be
obtained from the authors.
17The final expression is not presented here due to its complexity. Simulations reveal that for values of
qd and ∆ that secure positive demands for all goods in the initial stage the expression is strictly negative.
The binding constraint in the situation with no dicrimination is D3 > 0, which requires qd < 2651 and ∆ <
313 .
Table 1. Trademark Protectionism: Pooled results
(1)a (2)a (3)a (4)a (5)a (6)a
Difference in quality -0.16??? -0.19??? -0.15??? -0.16??? -0.19??? -0.15???
(∆qi,c) (0.05) (0.03) (0.05) (0.05) (0.05) (-0.04)
Import share -0.01 -0.02 -0.05
(mc) (0.07) (0.08) (0.07)
Constant -0.06?? -0.02 -0.08? -0.05? -0.02 -0.12?
(0.03) (0.06) (0.05) (0.03) (0.05) (0.07)
Industry dummy No Yes Yes No Yes Yes
Destination country dummy No No Yes No No Yes
R2-adjusted 0.01 0.02 0.05 0.01 0.02 0.05
# observations 1560 1560 1560 1560 1560 1560
aThe endogenous variable is given by the registration discrimination index against foreign firms, di,c. Allregressions are estimated using ordinary least squares, except for those where mc enters as an explanatoryvariable, in which cases we have used instrumental variables to correct for the potential endogeneity bias.Figures in parenthesis are White-robust standard errors. ??? stands for significance at the 1 percent level;?? for significance at the 5 percent level, and ? for significance at the 10 percent level.
28
Table 2. Trademark Protectionism by Country
Chinaa Chinaa Hong Hong Indiaa Indiaa South SouthKonga Konga Africaa Africaa
Difference in quality 0.21? 0.04 -0.17?? -0.12 -0.21??? -0.20?? -0.29??? -0.44??
(∆qi,c) (0.11) (0.17) (0.08) (0.10) (0.08) (0.09) (0.10) (0.13)
Constant -0.29??? -0.26??? -0.13??? -0.29??? 0.02 0.09 0.15??? 0.25???
(0.05) (0.18) (0.04) (0.15) (0.03) (0.07) (0.05) (0.08)
Industry dummy No Yes No Yes No Yes No YesR2-adjusted 0.01 0.05 0.01 0.10 0.02 0.12 0.01 0.08# observations 421 421 407 407 361 361 371 371
aThe endogenous variable is given by the registration discrimination index against foreign firms in eachof the four countries, di,c. All regressions are estimated using ordinary least squares. Figures in parenthesisare White-robust standard errors. ??? stands for significance at the 1 percent level; ?? for significance atthe 5 percent level, and ? for significance at the 10 percent level.
Table 3. Trademark Protectionism by Country: does import volume matter?
Chinaa Hong Konga Indiaa South Africaa
Difference in quality 0.02 -0.03 -0.31??? -0.44???
(∆qi,c) (0.17) (0.12) (0.11) (0.07)
Import share 0.41??? 0.14??? 0.74??? 0.54
(mc) (0.13) (0.04) (0.29) (0.41)
Constant -0.24 -0.39?? 0.08 0.38???
(0.16) (0.17) (0.08) (0.13)
Industry dummy Yes Yes Yes Yes
R2-adjusted 0.06 0.10 0.12 0.08
# observations 421 407 361 371
aThe endogenous variable is given by the registration discrimination index against foreign firms in eachof the four countries, di,c. All regressions are estimated instrumental variables to correct for the potentialendogeneity bias of mc. Figures in parenthesis are White-robust standard errors. ??? stands for significanceat the 1 percent level; ?? for significance at the 5 percent level, and ? for significance at the 10 percentlevel.